Concerning your 2 questions I have no answer right now. I am using the ARMKeil compiler and it silently compiles all without warnings. What you could check is, if TRICE("Hello, world\n"); works well in your environment. In any case TRICE0("Hello, TRICE\n"); should work. I hope to find some time in the next days to play around with gcc and maybe Linux too to get a clue.

What I am wondering too is, that only mode 0 works for you. The other modes have nothing to do with hardware despite TRICE_ENTER_CRITICAL_SECTION and TRICE_LEAVE_CRITICAL_SECTION what you can set empty if trice not using inside interrupts. Of course you need to have an eye on the buffer depth.

Thank you for the feedback.
Indeed, TRICE0("hello") works fine. It seems to be the argument counting macro COUNT_ARGUMENTS that doesn't count to zero correctly. (Redefining it as #define COUNT_ARGUMENTS(...) 0 would make the TRICE("hello") work). My compiler is GCC 8 if that turns out to be a compiler bug. I'll try to find time to dig deeper into this.

I managed to get the mode 200 working too. But doing a Trice log in the systick handler causes the log stream to freeze.
Should this work, if the TRICE_ENTER_CRITICAL_SECTION is set up correctly? (even leaving it defined to just a { causes the freeze, so I am worried I'm abusing this somehow).

void sys_tick_handler(void)
{
	system_millis++;
	#if TRICE_MODE == 200
	TRICE("systick happened at system_millis: %d\n", system_millis);
	triceTriggerTransmit();
	#endif
}

You are the only one reporting such issues. Do you have the chance to change the compiler, just for checking the COUNT_ARGUMENT issue?

The critical sections are essential for trice usage inside interrupts. But with empty macros you should see a result, even the data stream could get corrupted. Could it be, you produce too much data? You could try to use TRICE( "hi" ); only every 100ms inside the systick. Does this work?